AI is removing the middle class of software engineering?
AI coding tools are letting even weak developers generate large volumes of seemingly working code, raising fears of unmaintainable systems, hidden bugs and faster accumulation of technical debt. Many engineers report that senior reviewers and sound engineering culture have become bottlenecks, as they struggle to understand and safely approve AI-generated changes. At the same time, commenters worry that entry- and mid-level software roles are being squeezed, with productivity gains flowing mainly to a small number of strong engineers and to management cost-cutting, potentially hollowing out the career ladder into senior positions.
Impact of AI on code quality and architecture
- Many see LLMs as turbocharging “Big Ball of Mud” code: huge, brittle, poorly structured codebases that “work” in demos but fail under scale, change, or scrutiny.
- AI makes it cheap to generate tens of thousands of lines that compile, pass superficial tests, and look okay—until maintenance, performance, security, and migrations become nightmares.
- Several note this was always a problem with humans; AI mainly accelerates it and spreads it to more people and orgs, especially those with weak engineering culture.
Amplifying good vs bad engineers
- Strong consensus that AI is a force multiplier: good engineers get faster; weak ones cause far more damage, far more quickly.
- “Bad” here often means: no architectural sense, no ownership, no attention to invariants, backpressure, performance, or complexity.
- A single weak engineer with AI can flood code review with slop and overwhelm the people who actually understand the system.
Process, review, and management dynamics
- Traditional safeguards (small PRs, code review, tests, QA) weren’t designed for AI-scale throughput; reviewers become bottlenecks.
- Some companies are explicitly rewarding AI usage (token spend, AI % of work, PR count), incentivizing volume over understanding.
- Good practice with AI is described as: constrain agents, keep changes small, keep a human mental model, and invest heavily in review and tooling. Many orgs are doing the opposite.
Jobs, “middle class” SWE, and pipeline
- Strong concern that AI plus offshoring and oversupply are hollowing out junior/mid-level roles, making it harder to grow future seniors.
- Some argue the field is becoming winner‑take‑all: a few exceptional people plus AI can replace larger teams; others say demand for software is still huge and work will shift, not vanish.
- Disagreement on causality: some blame LLMs, others interest rates, hiring practices, immigration, and long‑standing lack of mentorship.
How people are actually using AI
- Many senior engineers describe using AI heavily but under tight control: for boilerplate, refactors, tests, and design exploration, not as an autonomous coder.
- Others report management mandating “AI first” everywhere, leading to vibe‑coded products no one fully understands.
Longer-term worries
- Fears of skill atrophy (“cognitive surrender”) and loss of understanding as people outsource thinking to agents.
- Some see this as part of a broader trend toward consolidation, technofeudalism, and a collapsing professional middle class; others think it’s another tooling shift that will stabilize once practices catch up.